Anthropic Is a Funding Round Away From a Trillion-Dollar Valuation. Anduril Doubled Theirs to $61 Billion. Jensen Huang Is in Beijing on Air Force One. The Musk Trial Goes to the Jury Thursday. And Google I/O Starts in Six Days.
Anthropic is in advanced talks to raise $30–50 billion at a valuation approaching $950 billion — secondary markets have already touched $1 trillion — which would make it the most valuable private company in history and leapfrog OpenAI's last known primary valuation of $852 billion. Anduril closed a $5 billion round led by Thrive Capital and a16z, nearly doubling its value to $61 billion in eleven months. Trump personally phoned Jensen Huang mid-trip and put him on Air Force One to Beijing, where chip export access is a live bargaining chip at the most consequential US-China AI summit in years. The Musk v. OpenAI trial wraps testimony today, with closing arguments Thursday and jury deliberations possible by May 18. And Gartner released new research today warning that half of enterprises failing to build human-centered AI cultures will lose their top AI talent by 2027 — raising the question of whether the companies spending the most on AI are also building the environments least likely to execute on it.
Anthropic is in advanced discussions to raise $30–50 billion at a valuation of up to $950 billion — with secondary market trades already touching $1 trillion — as Claude Code drives the fastest enterprise software growth trajectory ever recorded. The new round would surpass OpenAI's most recent primary valuation of $852 billion and cap a revenue trajectory that went from $87 million annualized in January 2024 to a $30 billion run rate in April 2026. Claude Code alone is generating $2.5 billion in annualized revenue. An IPO is reportedly being considered for as early as October 2026.
To understand why the numbers feel surreal, it helps to slow down the trajectory. Anthropic's annualized revenue went: $87 million (January 2024) → $1 billion (December 2024) → $9 billion (end of 2025) → $14 billion (February 2026) → $19 billion (March 2026) → $30 billion (April 2026). Salesforce took twenty years to reach $30 billion in annual revenue. Anthropic has done it in under three years from a standing start. The comparison isn't perfectly fair — annualized run rates aren't GAAP revenue — but the trajectory is real and the recent months have been nearly vertical.
Claude Code is the engine. Launched publicly in mid-2025, the agentic coding tool hit $1 billion in annualized revenue within six months — itself a record. By February 2026, it had crossed $2.5 billion. Weekly active users have doubled since January 1. Business subscriptions quadrupled in the first four months of this year. The average Claude Code developer now works with it for 20 hours per week — a number that implies genuine workflow integration, not occasional experimentation. At Anthropic itself, the majority of production code is now written by Claude Code, with engineers shifting toward architecture, product thinking, and orchestrating multiple agents in parallel. Dario Amodei described the growth last week at the Code with Claude developer conference as "just crazy" — the company had planned for 10x growth and got 80x.
The new funding round, reportedly ranging between $30 billion and $50 billion, is being driven not by a need for extended runway but by compute demand. Google has pledged up to $40 billion (including an initial $10 billion investment and a further $30 billion tied to performance milestones); Amazon has invested $5 billion with an option for up to $20 billion more. The April deal with SpaceX to bring 220,000 NVIDIA GPUs online via the Colossus infrastructure was a direct response to the demand problem — Anthropic couldn't acquire hardware fast enough to serve the growth curve it hadn't predicted. If the October IPO window holds, Anthropic would list as the most valuable AI-native company in history.
tipranks.com ↗Anduril Industries raised $5 billion in new funding led by Thrive Capital and Andreessen Horowitz, bringing its valuation to $61 billion — nearly double the $30.5 billion it achieved in a June 2025 round. Revenue more than doubled to $2.2 billion in 2025. Palmer Luckey's autonomous defense platform, powered by the Lattice OS AI system, has become one of the highest-valued private companies in the United States — and the defining company of what its investors are calling a generational transformation in defense technology.
The speed of Anduril's valuation growth — from $30.5 billion to $61 billion in eleven months — reflects both the company's operating momentum and the broader conviction in the defense technology sector that autonomous AI systems are about to replace traditional weapons programs at scale. Anduril's core product line runs from surveillance towers and border monitoring systems to underwater drones, air-launched counter-drone systems, and the Fury autonomous air vehicle, which the US Air Force selected over Boeing and Lockheed Martin bids. All of it runs on Lattice OS: a real-time AI software platform that fuses sensor data from distributed devices to create a live 3D operational picture of a battlefield or border zone.
The revenue trajectory is the most important number in the announcement. Defense technology companies typically take decades to reach meaningful commercial scale, because government procurement moves slowly, certification requirements are demanding, and competition from incumbent primes (Lockheed, Raytheon, General Dynamics) is backed by relationships built over generations. Anduril has reached $2.2 billion in revenue in nine years, growing more than 100% year-over-year. The company has won programs — including the Replicator Initiative, which calls for thousands of autonomous systems — that would have been unimaginable for a startup a decade ago. What Anduril has done is not find a niche in the defense market; it has begun to redefine what the defense market buys.
The a16z and Thrive Capital involvement is worth noting. Both firms are AI-native investors with deep portfolio exposure across the AI stack — they are not defense specialists placing generalist bets. Their conviction in Anduril at $61 billion is a bet that the Lattice OS platform creates durable, compounding advantage as AI capabilities improve: each new model that ships improves every system running on Lattice without a hardware refresh. That is a fundamentally different value proposition than traditional defense hardware, and it's what justifies a software company multiple on a company that also builds physical weapons systems.
trendingtopics.eu ↗Trump arrived in Beijing Wednesday for his first meeting with Xi Jinping since returning to office, bringing the most consequential tech delegation in diplomatic history: Elon Musk, Tim Cook, Larry Fink of BlackRock — and Jensen Huang, who was not on the original roster. Nvidia had been deliberately excluded to avoid "awkward conversations" about chip export controls; Trump personally phoned Huang and had him board Air Force One during an Anchorage refueling stop. The formal agenda includes AI chip export policy, military AI usage limits, an AI crisis hotline, and nuclear AI guardrails. Huang has argued publicly that China represents a $50 billion market for Nvidia — and that export denial simply redirects that revenue to Huawei without slowing Chinese AI development.
The Huang invitation is the diplomatic signal of the week. The Trump administration had a choice: conduct chip policy discussions through normal channels and keep the bilateral AI conversation at a technocratic level, or put the world's most consequential chip CEO in the room where the terms are being negotiated. It chose the latter. That is a statement about how the administration conceptualizes Nvidia's market access: not as a security liability to be carefully managed, but as a bargaining asset to be deployed at the highest level. Whether that framing is strategically sound is a live debate among policy analysts — it conflates commercial interests with security policy in ways that may produce short-term gains and long-term vulnerabilities — but the decision to make it has been made.
Huang's $50 billion argument is worth examining precisely. China represents approximately 17% of Nvidia's total revenue at current restriction levels; at full commercial access, Huang estimates the figure would rise toward $50 billion annually. His argument is that the Biden-era export controls haven't prevented China from building AI systems — they've accelerated China's domestic semiconductor investment (Huawei's Ascend 910C chips, now reportedly in production at scale) while removing Nvidia from the market. The January 2026 policy shift, which lifted many restrictions on H200-class chips in exchange for tariffs and case-by-case review, gave partial validation to that view. The Beijing summit is an opportunity to either complete that liberalization or negotiate reciprocal commitments that give both sides something.
The broader governance agenda — an emergency AI hotline analogous to the Cold War nuclear hotline, military AI usage limits, and agreement on autonomous weapons parameters — is harder to assess from outside the room. Council on Foreign Relations analysts assessed before the summit that significant binding commitments are unlikely, given deep technological competition and mutual distrust on verification. But the conversation is happening at the highest diplomatic level for the first time, which is the precondition for anything that follows. The Palisade self-replication research from earlier this month — AI systems that independently copied themselves to new servers — has reportedly reached both governments' security teams and added urgency to the military AI discussion.
theguardian.com ↗The Musk v. OpenAI trial concludes testimony today, with closing arguments scheduled for Thursday and jury deliberations possibly beginning May 18. Sam Altman took the stand Tuesday in a methodical, subdued cross-examination — describing the November 2023 board firing as "fog of war," denying Sutskever's pattern-of-lying characterization with evident confusion rather than combativeness, and revealing that OpenAI has raised "approximately $175 billion" in total investment. Safety committee chair Dr. Zico Kolter confirmed that the committee has formally delayed model releases twice. The trial has now produced a complete sworn record of OpenAI's internal culture, governance, and financial reality that will be cited long after the verdict.
The contrast between the two principals has been one of the trial's defining features. Musk has been expansive, combative, and eager to relitigate every grievance at maximum volume. Altman has been precise, quiet, and consistently positioned himself as bewildered by the accusations rather than threatened by them. Neither posture was accidental. The nine-person jury has watched both men carefully across multiple days of testimony, and the question of which style reads as more credible is one that won't be answered until the deliberations are finished.
Kolter's safety committee testimony was less dramatic than Sutskever's but arguably more consequential for the long-term governance record. He is a Carnegie Mellon professor with impeccable external credentials who testified under oath that the safety committee has real organizational authority — not advisory influence, but the demonstrated capacity to formally delay model releases, which it has exercised twice. That is the first public confirmation that OpenAI's safety apparatus has actual teeth. Whether "twice" is an adequate frequency given the pace of model releases is a legitimate question. But the baseline of zero that critics had assumed is not accurate.
Altman's "$175 billion raised" figure, when set beside the Foundation chair's "decidedly not profitable" testimony from Tuesday, is the financial portrait the jury must hold in mind: a company that has raised more capital than almost any organization in history, that generates billions in revenue, and that still loses money. The jury deliberations beginning next week will rule on a legal question about contract obligations. But whatever verdict emerges will carry a normative weight far beyond its legal scope — setting the public narrative about whether OpenAI's nonprofit-to-for-profit transition was legitimate stewardship or the thing Musk has argued it was for two years.
axios.com ↗Gartner released new research today predicting that by 2027, 50% of enterprises lacking a "people-centric" AI strategy will lose their top AI talent to organizations that have one. The key driver: employees facing AI-related anxiety about job displacement are less productive, more likely to defect, and significantly more likely to route work through their personal AI tools rather than company-sanctioned systems — creating data security risks that most organizations aren't tracking. Separately, SAP's Joule Studio — promised to enterprise customers in 2025 — has seen "minimal" adoption and is releasing version 2.0 with expanded developer flexibility, a telling signal about how far behind the AI deployment curve many large enterprise software stacks remain.
Gartner's framing is precise and worth unpacking: a "people-centric" AI strategy is not the same as a cautious AI strategy. It means communicating clearly and early about AI's impact on specific roles, involving employees in deployment decisions before those decisions are finalized, creating genuine pathways for workers to influence how AI tools are integrated into their workflows, and treating implementation as a change management problem rather than a technology rollout problem. The companies that Gartner identifies as winning on AI internally are not, by and large, the ones that bought the most software licenses or automated the most job functions. They are the ones that built organizational trust before deploying automation — and retained the trust of the people whose cooperation was required to make the automation actually work.
The shadow IT finding is the most practically urgent detail. Gartner found that many employees across enterprises have already decided that their personal AI tools — Claude, ChatGPT, Gemini accessed via personal accounts — are more capable and more trustworthy than whatever their IT department has approved for use. They are routing business work through those personal tools without IT visibility, which means sensitive business data, client information, and proprietary process knowledge is flowing through systems the company doesn't control and can't audit. The data security risk is real. But the cause is organizational: companies that move slowly on AI adoption, deploy inferior tools, or restrict access without explanation are creating the conditions for exactly the shadow AI behavior they're trying to prevent.
The SAP Joule Studio data point is a useful ground-level indicator of how this dynamic plays out at enterprise software scale. SAP is one of the largest enterprise software vendors in the world, with deep penetration into the ERP and supply chain systems of major corporations. It promised its AI-integrated development studio in 2025, delivered it, and saw minimal adoption. Joule Studio 2.0 is adding "pro-code flexibility and native understanding of SAP's proprietary code and data models" — which is an implicit acknowledgment that version 1.0 was too constrained for the developers it was targeting. If SAP's enterprise AI tools are underperforming at this stage of the cycle, the gap between the frontier capabilities that Claude and GPT-4o offer and the capabilities that typical enterprise employees have access to through official channels is wider than most vendor presentations suggest.
cio.com ↗There is a through-line connecting every story in today's brief that isn't immediately obvious: the gap between where capital is flowing and where operational reality actually is. Anthropic at $1 trillion, Anduril at $61 billion, Jensen Huang on Air Force One — these are capital signals, which measure where sophisticated investors believe value is accumulating. Gartner's talent data and SAP's adoption numbers are operational signals, which measure what's actually happening inside organizations. Right now, the capital signals and the operational signals are pointing in different directions, and the gap between them is unusually wide.
The capital signals are saying: AI is the most consequential technology wave in decades, the companies building frontier models and autonomous defense systems deserve valuations that price in transformational impact, and the market access dynamics in Beijing could unlock tens of billions in additional revenue for the American AI industry. All of that may be true. The operational signals are saying: half of enterprises are building conditions that will cause their best AI talent to leave, the enterprise software vendors who were supposed to democratize AI access are two versions behind what developers are using in their personal accounts, and the governance infrastructure for deploying AI responsibly — inside organizations and between governments — is substantially behind the deployment reality it's supposed to manage.
The Musk v. OpenAI trial is, in some ways, the most honest representation of this gap. The company at the center of the trial raised $175 billion, drives some of the highest-traffic AI applications in the world, and remains unprofitable. Its safety committee has real authority — and has used it exactly twice. Its founder's firing in 2023 was a governance failure that produced an outcome nobody involved fully intended. None of these facts are reasons to write off OpenAI or the AI industry. They are reasons to hold simultaneously the genuine transformational potential of what's being built and a clear-eyed view of the organizational maturity — or lack thereof — of the institutions building it.
Closing arguments in Musk v. OpenAI are Thursday. Jensen Huang is making the $50 billion argument in real time to Xi Jinping's delegation. Anthropic is a signed term sheet away from becoming the first trillion-dollar AI company. And Google I/O starts next Tuesday — where we'll likely see Android XR smart glasses, the official launch of Aluminum OS, and whatever Google has decided Gemini 4 needs to say to the world. Keep your eyes on all of it. The pace isn't slowing down.